Re:Think 1 October, 2025

Labour Party conference: Frontier AI with Kanishka Narayan MP

Joe Hill
Directory of Strategy

Politics thrives on the day-to-day cycle of announcements, scandals, successes and failures. But sometimes that means Westminster loses sight of the big arcs of history, which can be happening all around us but pass unnoticed.

AI is the defining technological shift of our generation. The people who didn’t believe it would change the world, as recently as four years ago, have all been proven wrong. The future capabilities of AI are hard to predict, but policymakers would be foolish to bet against further radical transformation powered by generative AI.

It's this context that we sat down to talk to Kanishka Narayan MP, the recently appointed Minister for AI and Online Safety, about how he sees Britain’s future in a world transformed by AI. Here are my three key reflections on our discussion:

 

  1. Sovereignty won’t be universal.

Policymakers are increasingly keen for countries to have their own stake in the AI race, but calls for blanket sovereignty are unrealistic. Particularly for a trading nation like the UK, with a large services sector, there are few supply chains which we depend on and can control domestically end-to-end – our economy is based on imports and exports. In reality, no country today has total sovereign control over its supply chain. For the UK, the goal of sovereignty should be about the ability to make our own decisions, which starts with a capable government which understands frontier developments in technology. Beyond that, we need an AI policy which can double-down on our strengths in key areas to make sure we control the right capabilities to have a seat at the table. Sectors, like the use of AI in defence and advanced manufacturing, can be our comparative advantages, as can some core capabilities for model design and testing which can only be controlled by companies with access to highly mobile global talent.

  1. We’re balancing different kinds of risk.

There are plenty of risks involved in the development of frontier AI, and in widespread adoption through the economy. Safety risks are hard to mitigate without international agreements, which historically the UK has been a leader in. And many of the social and economic impacts that widespread AI will have are still unknown. But the risk where we miss out on this wave of transformation, like the UK has missed out on many of the benefits of the previous software revolutions, are also great. It’s important for policymakers to avoid the backlash that could mean AI adoption is slowed or stopped in the UK, which would be disastrous in a global economy where others would forge ahead in our place. We wrote about the impact of this in public services in our paper Getting the machine learning from last year.

  1. Good adoption will be bottom-up.

Previous software revolutions, like the adoption of cloud-based software-as-a-service, have been bottom-up and driven by individual customers promoting the technology they like to use within their organisations. In the public sector in particular, our policy approach tends to prescribe top-down and centralising instincts, without recognising the huge value that letting people experiment and learn comes with. When Kanishka was appointed we wrote an open letter to him and Liz Kendall about how to create the environment for this in government.